Update README.md

README.md

@@ -18,22 +18,23 @@ Here are the specifications of the process used:
 - The MG-Verilog dataset was downloaded from Gatech-EIC/MG-Verilog. I have copied this dataset to my GitHub repository used to train the model.
 - The dataset and the training script was uploaded to my Google Drive for easy access on Colab.
 - A 4-bit quantized version of the base model is loaded using a quantization configuration using BitsAndBytesConfig and AutoModelForCausalLM.from_pretrained. The follwing are the specifications of the quantization configuration:
-
+  ```
   load_in_4bit=True,
   bnb_4bit_use_double_quant=True,
   bnb_4bit_quant_type="nf4",
   bnb_4bit_compute_dtype=torch.bfloat16
+  ```
   
 - The model is prepared for k-bit training using prepare_model_for_kbit_training.
 - The PEFT configuartion is set and a PeftModel is created from the quantized model. Here are the specifications of the PEFT configuration:
-
+  ```
   r=64,
   lora_alpha=16,
   target_modules=["q_proj", "v_proj"],
   lora_dropout=0.05,
   bias="none",
   task_type=TaskType.CAUSAL_LM
-
+  ```
 - The Data Collator class implemented in the training script has been taken directly from qlora.py from the MG-Verilog GitHub (https://github.com/GATECH-EIC/mg-verilog)
 
 ## Citations